Aeon Flux based on the MTV Liquid Television animation series.
Description
Use weight 0.6 - 0.8. Higher weights will cause distortion.
You may have to inpaint the eyes/face in full body shots. It has some problems with her curly hair tips. The faces in the first two example images were inpainted, the rest are raw output. I tried to also get it to re-create the animation style of the original series. You can try using the tag "animation" to make the style stronger.
Simple prompts work better. Example prompt:
aeon flux, black hair, blue eyes, half body, interior, futuristic city
FAQ
Comments (6)
Use weight 0.6 - 0.8. Higher weights will cause distortion.
You may have to inpaint the eyes/face in full body shots. It has some problems with her curly hair tips. The faces in the first two example images were inpainted, the rest are raw output. I tried to also get it to re-create the animation style of the original series. You can try using the tag "animation" to make the style stronger.
Simple prompts work better. Example prompt:
aeon flux, black hair, blue eyes, half body, interior, f
I can't believe it! Just the day before yesterday, I was testing loras for anime and comics and wondering how is it possible that no one has trained a model for Aeon Flux yet!
This one was kind of difficult, I re-trained it a few times. I also tried to make one that was more character focused and didn't bring in the style. I might re-do that one. I'll probably upload a version 2 at some point.
I was blown away by the versatility of your Lora and its ability to produce high-quality images across different styles, from the classic animated series to semi-realistic and even photorealistic representations of the character. Your dedication and hard work in perfecting this model are truly impressive. I spent an entire afternoon generating images because I was so excited about its capabilities haha. Thank you so much for sharing your work!
Thanks for the feedback, glad you were able to get some great stuff out of it!
This is such an incredibly well trained Lora 👏! Would love to know your workflow or your thoughts on what leads to success when training. E.g. number of images, things to have in captions, training parameters, was it done using kohya?